}
unsigned AArch64FastISel::fastMaterializeAlloca(const AllocaInst *AI) {
- assert(TLI.getValueType(AI->getType(), true) == MVT::i64 &&
+ assert(TLI.getValueType(DL, AI->getType(), true) == MVT::i64 &&
"Alloca should always return a pointer.");
// Don't handle dynamic allocas.
unsigned char OpFlags = Subtarget->ClassifyGlobalReference(GV, TM);
- EVT DestEVT = TLI.getValueType(GV->getType(), true);
+ EVT DestEVT = TLI.getValueType(DL, GV->getType(), true);
if (!DestEVT.isSimple())
return 0;
}
unsigned AArch64FastISel::fastMaterializeConstant(const Constant *C) {
- EVT CEVT = TLI.getValueType(C->getType(), true);
+ EVT CEVT = TLI.getValueType(DL, C->getType(), true);
// Only handle simple types.
if (!CEVT.isSimple())
U = C;
}
- if (
const PointerType *Ty = dyn_cast<PointerType>(Obj->getType()))
+ if (
auto *Ty = dyn_cast<PointerType>(Obj->getType()))
if (Ty->getAddressSpace() > 255)
// Fast instruction selection doesn't support the special
// address spaces.
}
case Instruction::IntToPtr: {
// Look past no-op inttoptrs.
- if (TLI.getValueType(U->getOperand(0)->getType()) == TLI.getPointerTy())
+ if (TLI.getValueType(DL, U->getOperand(0)->getType()) ==
+ TLI.getPointerTy(DL))
return computeAddress(U->getOperand(0), Addr, Ty);
break;
}
case Instruction::PtrToInt: {
// Look past no-op ptrtoints.
- if (TLI.getValueType(U->getType()) == TLI.getPointerTy())
+ if (TLI.getValueType(DL, U->getType()) == TLI.getPointerTy(DL))
return computeAddress(U->getOperand(0), Addr, Ty);
break;
}
case Instruction::IntToPtr:
// Look past no-op inttoptrs if its operand is in the same BB.
if (InMBB &&
- TLI.getValueType(U->getOperand(0)->getType()) == TLI.getPointerTy())
+ TLI.getValueType(DL, U->getOperand(0)->getType()) ==
+ TLI.getPointerTy(DL))
return computeCallAddress(U->getOperand(0), Addr);
break;
case Instruction::PtrToInt:
// Look past no-op ptrtoints if its operand is in the same BB.
- if (InMBB &&
- TLI.getValueType(U->getType()) == TLI.getPointerTy())
+ if (InMBB && TLI.getValueType(DL, U->getType()) == TLI.getPointerTy(DL))
return computeCallAddress(U->getOperand(0), Addr);
break;
}
bool AArch64FastISel::isTypeLegal(Type *Ty, MVT &VT) {
- EVT evt = TLI.getValueType(Ty, true);
+ EVT evt = TLI.getValueType(DL, Ty, true);
// Only handle simple types.
if (evt == MVT::Other || !evt.isSimple())
// Cannot encode an offset register and an immediate offset in the same
// instruction. Fold the immediate offset into the load/store instruction and
- // emit an additonal add to take care of the offset register.
+ // emit an additional add to take care of the offset register.
if (!ImmediateOffsetNeedsLowering && Addr.getOffset() && Addr.getOffsetReg())
RegisterOffsetNeedsLowering = true;
// FIXME: We shouldn't be using getObjectSize/getObjectAlignment. The size
// and alignment should be based on the VT.
MMO = FuncInfo.MF->getMachineMemOperand(
- MachinePointerInfo::getFixedStack(FI, Offset), Flags,
- MFI.getObjectSize(FI), MFI.getObjectAlignment(FI));
+ MachinePointerInfo::getFixedStack(*FuncInfo.MF, FI, Offset), Flags,
+ MFI.getObjectSize(FI), MFI.getObjectAlignment(FI));
// Now add the rest of the operands.
MIB.addFrameIndex(FI).addImm(Offset);
} else {
}
// Check if the mul can be folded into the instruction.
- if (RHS->hasOneUse() && isValueAvailable(RHS))
+ if (RHS->hasOneUse() && isValueAvailable(RHS)) {
if (isMulPowOf2(RHS)) {
const Value *MulLHS = cast<MulOperator>(RHS)->getOperand(0);
const Value *MulRHS = cast<MulOperator>(RHS)->getOperand(1);
if (!RHSReg)
return 0;
bool RHSIsKill = hasTrivialKill(MulLHS);
- return emitAddSub_rs(UseAdd, RetVT, LHSReg, LHSIsKill, RHSReg, RHSIsKill,
- AArch64_AM::LSL, ShiftVal, SetFlags, WantResult);
+ ResultReg = emitAddSub_rs(UseAdd, RetVT, LHSReg, LHSIsKill, RHSReg,
+ RHSIsKill, AArch64_AM::LSL, ShiftVal, SetFlags,
+ WantResult);
+ if (ResultReg)
+ return ResultReg;
}
+ }
// Check if the shift can be folded into the instruction.
- if (RHS->hasOneUse() && isValueAvailable(RHS))
+ if (RHS->hasOneUse() && isValueAvailable(RHS)) {
if (const auto *SI = dyn_cast<BinaryOperator>(RHS)) {
if (const auto *C = dyn_cast<ConstantInt>(SI->getOperand(1))) {
AArch64_AM::ShiftExtendType ShiftType = AArch64_AM::InvalidShiftExtend;
if (!RHSReg)
return 0;
bool RHSIsKill = hasTrivialKill(SI->getOperand(0));
- return emitAddSub_rs(UseAdd, RetVT, LHSReg, LHSIsKill, RHSReg,
- RHSIsKill, ShiftType, ShiftVal, SetFlags,
- WantResult);
+ ResultReg = emitAddSub_rs(UseAdd, RetVT, LHSReg, LHSIsKill, RHSReg,
+ RHSIsKill, ShiftType, ShiftVal, SetFlags,
+ WantResult);
+ if (ResultReg)
+ return ResultReg;
}
}
}
+ }
unsigned RHSReg = getRegForValue(RHS);
if (!RHSReg)
if (RetVT != MVT::i32 && RetVT != MVT::i64)
return 0;
+ // Don't deal with undefined shifts.
+ if (ShiftImm >= RetVT.getSizeInBits())
+ return 0;
+
static const unsigned OpcTable[2][2][2] = {
{ { AArch64::SUBWrs, AArch64::SUBXrs },
{ AArch64::ADDWrs, AArch64::ADDXrs } },
if (RetVT != MVT::i32 && RetVT != MVT::i64)
return 0;
+ if (ShiftImm >= 4)
+ return 0;
+
static const unsigned OpcTable[2][2][2] = {
{ { AArch64::SUBWrx, AArch64::SUBXrx },
{ AArch64::ADDWrx, AArch64::ADDXrx } },
bool AArch64FastISel::emitCmp(const Value *LHS, const Value *RHS, bool IsZExt) {
Type *Ty = LHS->getType();
- EVT EVT = TLI.getValueType(Ty, true);
+ EVT EVT = TLI.getValueType(DL, Ty, true);
if (!EVT.isSimple())
return false;
MVT VT = EVT.getSimpleVT();
return ResultReg;
// Check if the mul can be folded into the instruction.
- if (RHS->hasOneUse() && isValueAvailable(RHS))
+ if (RHS->hasOneUse() && isValueAvailable(RHS)) {
if (isMulPowOf2(RHS)) {
const Value *MulLHS = cast<MulOperator>(RHS)->getOperand(0);
const Value *MulRHS = cast<MulOperator>(RHS)->getOperand(1);
if (!RHSReg)
return 0;
bool RHSIsKill = hasTrivialKill(MulLHS);
- return emitLogicalOp_rs(ISDOpc, RetVT, LHSReg, LHSIsKill, RHSReg,
- RHSIsKill, ShiftVal);
+ ResultReg = emitLogicalOp_rs(ISDOpc, RetVT, LHSReg, LHSIsKill, RHSReg,
+ RHSIsKill, ShiftVal);
+ if (ResultReg)
+ return ResultReg;
}
+ }
// Check if the shift can be folded into the instruction.
- if (RHS->hasOneUse() && isValueAvailable(RHS))
+ if (RHS->hasOneUse() && isValueAvailable(RHS)) {
if (const auto *SI = dyn_cast<ShlOperator>(RHS))
if (const auto *C = dyn_cast<ConstantInt>(SI->getOperand(1))) {
uint64_t ShiftVal = C->getZExtValue();
if (!RHSReg)
return 0;
bool RHSIsKill = hasTrivialKill(SI->getOperand(0));
- return emitLogicalOp_rs(ISDOpc, RetVT, LHSReg, LHSIsKill, RHSReg,
- RHSIsKill, ShiftVal);
+ ResultReg = emitLogicalOp_rs(ISDOpc, RetVT, LHSReg, LHSIsKill, RHSReg,
+ RHSIsKill, ShiftVal);
+ if (ResultReg)
+ return ResultReg;
}
+ }
unsigned RHSReg = getRegForValue(RHS);
if (!RHSReg)
{ AArch64::ORRWrs, AArch64::ORRXrs },
{ AArch64::EORWrs, AArch64::EORXrs }
};
+
+ // Don't deal with undefined shifts.
+ if (ShiftImm >= RetVT.getSizeInBits())
+ return 0;
+
const TargetRegisterClass *RC;
unsigned Opc;
switch (RetVT.SimpleTy) {
unsigned AArch64FastISel::emitLoad(MVT VT, MVT RetVT, Address Addr,
bool WantZExt, MachineMemOperand *MMO) {
- if(!TLI.allowsMisalignedMemoryAccesses(VT))
+ if (!TLI.allowsMisalignedMemoryAccesses(VT))
return 0;
// Simplify this down to something we can handle.
bool AArch64FastISel::emitStore(MVT VT, unsigned SrcReg, Address Addr,
MachineMemOperand *MMO) {
- if(!TLI.allowsMisalignedMemoryAccesses(VT))
+ if (!TLI.allowsMisalignedMemoryAccesses(VT))
return false;
// Simplify this down to something we can handle.
MIB.addImm(TestBit);
MIB.addMBB(TBB);
- // Obtain the branch weight and add the TrueBB to the successor list.
- uint32_t BranchWeight = 0;
- if (FuncInfo.BPI)
- BranchWeight = FuncInfo.BPI->getEdgeWeight(BI->getParent(),
- TBB->getBasicBlock());
- FuncInfo.MBB->addSuccessor(TBB, BranchWeight);
- fastEmitBranch(FBB, DbgLoc);
-
+ finishCondBranch(BI->getParent(), TBB, FBB);
return true;
}
MachineBasicBlock *TBB = FuncInfo.MBBMap[BI->getSuccessor(0)];
MachineBasicBlock *FBB = FuncInfo.MBBMap[BI->getSuccessor(1)];
- AArch64CC::CondCode CC = AArch64CC::NE;
if (const CmpInst *CI = dyn_cast<CmpInst>(BI->getCondition())) {
if (CI->hasOneUse() && isValueAvailable(CI)) {
// Try to optimize or fold the cmp.
// FCMP_UEQ and FCMP_ONE cannot be checked with a single branch
// instruction.
- CC = getCompareCC(Predicate);
+ AArch64CC::CondCode CC = getCompareCC(Predicate);
AArch64CC::CondCode ExtraCC = AArch64CC::AL;
switch (Predicate) {
default:
.addImm(CC)
.addMBB(TBB);
- // Obtain the branch weight and add the TrueBB to the successor list.
- uint32_t BranchWeight = 0;
- if (FuncInfo.BPI)
- BranchWeight = FuncInfo.BPI->getEdgeWeight(BI->getParent(),
- TBB->getBasicBlock());
- FuncInfo.MBB->addSuccessor(TBB, BranchWeight);
-
- fastEmitBranch(FBB, DbgLoc);
- return true;
- }
- } else if (TruncInst *TI = dyn_cast<TruncInst>(BI->getCondition())) {
- MVT SrcVT;
- if (TI->hasOneUse() && isValueAvailable(TI) &&
- isTypeSupported(TI->getOperand(0)->getType(), SrcVT)) {
- unsigned CondReg = getRegForValue(TI->getOperand(0));
- if (!CondReg)
- return false;
- bool CondIsKill = hasTrivialKill(TI->getOperand(0));
-
- // Issue an extract_subreg to get the lower 32-bits.
- if (SrcVT == MVT::i64) {
- CondReg = fastEmitInst_extractsubreg(MVT::i32, CondReg, CondIsKill,
- AArch64::sub_32);
- CondIsKill = true;
- }
-
- unsigned ANDReg = emitAnd_ri(MVT::i32, CondReg, CondIsKill, 1);
- assert(ANDReg && "Unexpected AND instruction emission failure.");
- emitICmp_ri(MVT::i32, ANDReg, /*IsKill=*/true, 0);
-
- if (FuncInfo.MBB->isLayoutSuccessor(TBB)) {
- std::swap(TBB, FBB);
- CC = AArch64CC::EQ;
- }
- BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::Bcc))
- .addImm(CC)
- .addMBB(TBB);
-
- // Obtain the branch weight and add the TrueBB to the successor list.
- uint32_t BranchWeight = 0;
- if (FuncInfo.BPI)
- BranchWeight = FuncInfo.BPI->getEdgeWeight(BI->getParent(),
- TBB->getBasicBlock());
- FuncInfo.MBB->addSuccessor(TBB, BranchWeight);
-
- fastEmitBranch(FBB, DbgLoc);
+ finishCondBranch(BI->getParent(), TBB, FBB);
return true;
}
} else if (const auto *CI = dyn_cast<ConstantInt>(BI->getCondition())) {
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::B))
.addMBB(Target);
- // Obtain the branch weight and add the target to the successor list.
- uint32_t BranchWeight = 0;
- if (FuncInfo.BPI)
- BranchWeight = FuncInfo.BPI->getEdgeWeight(BI->getParent(),
- Target->getBasicBlock());
- FuncInfo.MBB->addSuccessor(Target, BranchWeight);
+ // Obtain the branch probability and add the target to the successor list.
+ if (FuncInfo.BPI) {
+ auto BranchProbability = FuncInfo.BPI->getEdgeProbability(
+ BI->getParent(), Target->getBasicBlock());
+ FuncInfo.MBB->addSuccessor(Target, BranchProbability);
+ } else
+ FuncInfo.MBB->addSuccessorWithoutProb(Target);
return true;
- } else if (foldXALUIntrinsic(CC, I, BI->getCondition())) {
- // Fake request the condition, otherwise the intrinsic might be completely
- // optimized away.
- unsigned CondReg = getRegForValue(BI->getCondition());
- if (!CondReg)
- return false;
-
- // Emit the branch.
- BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::Bcc))
- .addImm(CC)
- .addMBB(TBB);
+ } else {
+ AArch64CC::CondCode CC = AArch64CC::NE;
+ if (foldXALUIntrinsic(CC, I, BI->getCondition())) {
+ // Fake request the condition, otherwise the intrinsic might be completely
+ // optimized away.
+ unsigned CondReg = getRegForValue(BI->getCondition());
+ if (!CondReg)
+ return false;
- // Obtain the branch weight and add the TrueBB to the successor list.
- uint32_t BranchWeight = 0;
- if (FuncInfo.BPI)
- BranchWeight = FuncInfo.BPI->getEdgeWeight(BI->getParent(),
- TBB->getBasicBlock());
- FuncInfo.MBB->addSuccessor(TBB, BranchWeight);
+ // Emit the branch.
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::Bcc))
+ .addImm(CC)
+ .addMBB(TBB);
- fastEmitBranch(FBB, DbgLoc);
- return true;
+ finishCondBranch(BI->getParent(), TBB, FBB);
+ return true;
+ }
}
unsigned CondReg = getRegForValue(BI->getCondition());
return false;
bool CondRegIsKill = hasTrivialKill(BI->getCondition());
- // We've been divorced from our compare! Our block was split, and
- // now our compare lives in a predecessor block. We musn't
- // re-compare here, as the children of the compare aren't guaranteed
- // live across the block boundary (we *could* check for this).
- // Regardless, the compare has been done in the predecessor block,
- // and it left a value for us in a virtual register. Ergo, we test
- // the one-bit value left in the virtual register.
- emitICmp_ri(MVT::i32, CondReg, CondRegIsKill, 0);
-
+ // i1 conditions come as i32 values, test the lowest bit with tb(n)z.
+ unsigned Opcode = AArch64::TBNZW;
if (FuncInfo.MBB->isLayoutSuccessor(TBB)) {
std::swap(TBB, FBB);
- CC = AArch64CC::EQ;
+ Opcode = AArch64::TBZW;
}
- BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, TII.get(AArch64::Bcc))
- .addImm(CC)
+ const MCInstrDesc &II = TII.get(Opcode);
+ unsigned ConstrainedCondReg
+ = constrainOperandRegClass(II, CondReg, II.getNumDefs());
+ BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II)
+ .addReg(ConstrainedCondReg, getKillRegState(CondRegIsKill))
+ .addImm(0)
.addMBB(TBB);
- // Obtain the branch weight and add the TrueBB to the successor list.
- uint32_t BranchWeight = 0;
- if (FuncInfo.BPI)
- BranchWeight = FuncInfo.BPI->getEdgeWeight(BI->getParent(),
- TBB->getBasicBlock());
- FuncInfo.MBB->addSuccessor(TBB, BranchWeight);
-
- fastEmitBranch(FBB, DbgLoc);
+ finishCondBranch(BI->getParent(), TBB, FBB);
return true;
}
BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc, II).addReg(AddrReg);
// Make sure the CFG is up-to-date.
- for (unsigned i = 0, e = BI->getNumSuccessors(); i != e; ++i)
- FuncInfo.MBB->addSuccessor(FuncInfo.MBBMap[BI->getSuccessor(i)]);
+ for (auto *Succ : BI->successors())
+ FuncInfo.MBB->addSuccessor(FuncInfo.MBBMap[Succ]);
return true;
}
bool AArch64FastISel::selectCmp(const Instruction *I) {
const CmpInst *CI = cast<CmpInst>(I);
+ // Vectors of i1 are weird: bail out.
+ if (CI->getType()->isVectorTy())
+ return false;
+
// Try to optimize or fold the cmp.
CmpInst::Predicate Predicate = optimizeCmpPredicate(CI);
unsigned ResultReg = 0;
if (SrcReg == 0)
return false;
- EVT SrcVT = TLI.getValueType(I->getOperand(0)->getType(), true);
+ EVT SrcVT = TLI.getValueType(DL, I->getOperand(0)->getType(), true);
if (SrcVT == MVT::f128)
return false;
return false;
bool SrcIsKill = hasTrivialKill(I->getOperand(0));
- EVT SrcVT = TLI.getValueType(I->getOperand(0)->getType(), true);
+ EVT SrcVT = TLI.getValueType(DL, I->getOperand(0)->getType(), true);
// Handle sign-extension.
if (SrcVT == MVT::i16 || SrcVT == MVT::i8 || SrcVT == MVT::i1) {
if (ArgTy->isStructTy() || ArgTy->isArrayTy())
return false;
- EVT ArgVT = TLI.getValueType(ArgTy);
+ EVT ArgVT = TLI.getValueType(DL, ArgTy);
if (!ArgVT.isSimple())
return false;
unsigned GPRIdx = 0;
unsigned FPRIdx = 0;
for (auto const &Arg : F->args()) {
- MVT VT = TLI.getSimpleValueType(Arg.getType());
+ MVT VT = TLI.getSimpleValueType(DL, Arg.getType());
unsigned SrcReg;
const TargetRegisterClass *RC;
if (VT >= MVT::i1 && VT <= MVT::i32) {
.addImm(NumBytes);
// Process the args.
- for (unsigned i = 0, e = ArgLocs.size(); i != e; ++i) {
- CCValAssign &VA = ArgLocs[i];
+ for (CCValAssign &VA : ArgLocs) {
const Value *ArgVal = CLI.OutVals[VA.getValNo()];
MVT ArgVT = OutVTs[VA.getValNo()];
unsigned Alignment = DL.getABITypeAlignment(ArgVal->getType());
MachineMemOperand *MMO = FuncInfo.MF->getMachineMemOperand(
- MachinePointerInfo::getStack(Addr.getOffset()),
- MachineMemOperand::MOStore, ArgVT.getStoreSize(), Alignment);
+ MachinePointerInfo::getStack(*FuncInfo.MF, Addr.getOffset()),
+ MachineMemOperand::MOStore, ArgVT.getStoreSize(), Alignment);
if (!emitStore(ArgVT, ArgReg, Addr, MMO))
return false;
return false;
// Make sure nothing is in the way
- BasicBlock::const_iterator Start = I;
- BasicBlock::const_iterator End = II;
+ BasicBlock::const_iterator Start(I);
+ BasicBlock::const_iterator End(II);
for (auto Itr = std::prev(Start); Itr != End; --Itr) {
// We only expect extractvalue instructions between the intrinsic and the
// instruction to be selected.
if (F.isVarArg())
return false;
+ if (TLI.supportSplitCSR(FuncInfo.MF))
+ return false;
+
// Build a list of return value registers.
SmallVector<unsigned, 4> RetRegs;
if (Ret->getNumOperands() > 0) {
CallingConv::ID CC = F.getCallingConv();
SmallVector<ISD::OutputArg, 4> Outs;
- GetReturnInfo(F.getReturnType(), F.getAttributes(), Outs, TLI);
+ GetReturnInfo(F.getReturnType(), F.getAttributes(), Outs, TLI, DL);
// Analyze operands of the call, assigning locations to each operand.
SmallVector<CCValAssign, 16> ValLocs;
if (!MRI.getRegClass(SrcReg)->contains(DestReg))
return false;
- EVT RVEVT = TLI.getValueType(RV->getType());
+ EVT RVEVT = TLI.getValueType(DL, RV->getType());
if (!RVEVT.isSimple())
return false;
MachineInstrBuilder MIB = BuildMI(*FuncInfo.MBB, FuncInfo.InsertPt, DbgLoc,
TII.get(AArch64::RET_ReallyLR));
- for (unsigned i = 0, e = RetRegs.size(); i != e; ++i)
- MIB.addReg(RetRegs[i], RegState::Implicit);
+ for (unsigned RetReg : RetRegs)
+ MIB.addReg(RetReg, RegState::Implicit);
return true;
}
Value *Op = I->getOperand(0);
Type *SrcTy = Op->getType();
- EVT SrcEVT = TLI.getValueType(SrcTy, true);
- EVT DestEVT = TLI.getValueType(DestTy, true);
+ EVT SrcEVT = TLI.getValueType(DL, SrcTy, true);
+ EVT DestEVT = TLI.getValueType(DL, DestTy, true);
if (!SrcEVT.isSimple())
return false;
if (!DestEVT.isSimple())
}
bool AArch64FastISel::selectRem(const Instruction *I, unsigned ISDOpcode) {
- EVT DestEVT = TLI.getValueType(I->getType(), true);
+ EVT DestEVT = TLI.getValueType(DL, I->getType(), true);
if (!DestEVT.isSimple())
return false;
bool IdxNIsKill = hasTrivialKill(Idx);
// If the index is smaller or larger than intptr_t, truncate or extend it.
- MVT PtrVT = TLI.getPointerTy();
+ MVT PtrVT = TLI.getPointerTy(DL);
EVT IdxVT = EVT::getEVT(Idx->getType(), /*HandleUnknown=*/false);
if (IdxVT.bitsLT(PtrVT)) {
IdxN = emitIntExt(IdxVT.getSimpleVT(), IdxN, PtrVT, /*IsZExt=*/false);
// into a single N = N + TotalOffset.
uint64_t TotalOffs = 0;
Type *Ty = I->getOperand(0)->getType();
- MVT VT = TLI.getPointerTy();
+ MVT VT = TLI.getPointerTy(DL);
for (auto OI = std::next(I->op_begin()), E = I->op_end(); OI != E; ++OI) {
const Value *Idx = *OI;
if (auto *StTy = dyn_cast<StructType>(Ty)) {
projects / oota-llvm.git / blobdiff